Update jolokia.go

[InfiniteCode]

Required for all PRs:

• CHANGELOG.md updated
• Sign CLA (if not already signed)
• README.md updated (if adding a new plugin)

Some of the values returned from jolokia have three levels of nesting. This fixes the issue of maps being returned instead of values.

[phemmer]

Holy nesting batman!

I think a programmatic solution would be better, not an explicit 3-key map.

For example:

func extractValues(k string, v interface{}, fields map[string]interface{}) {
    if v, ok := v.(map[string]interface{}); ok {
        for k2, v2 := range v {
            extractValues(k+"_"+k2, v2, fields)
        }
    } else {
        fields[k] = v
    }
}

func Gather() {
    ...
    if values, ok := out["value"]; ok {
        extractValues(measurement, values, fields)
    }
    ...
}

[InfiniteCode]

Totally agree. I don't think there ever will be more than 3 layers of nesting, but in a long term I think a recursive call is definitely a better solution.

We also need to be careful with recursion in general, in case there is some kind of issues might happen and recursion tree becomes too long. Probably limiting depth should be included.

[phemmer]

Well we're unmarshalling json, so cyclical structures aren't possible, so there's no risk of infinite recursion. You'd have to get thousands of levels deep before it would become an issue.
I guess what I'm saying is that I don't see the need for a depth limit. But the code complexity to add it is pretty minor, so I'm not really opposed to it either.

[sparrc]

I believe this will fix #1693 once merged.

@InfiniteCode do you think you could refactor the code to make a recursive call? I agree with @phemmer that there isn't much risk of infinite recursion when we're only talking about marshaling JSON.

[ScubaDrew]

Can we merge this as-is to get this working until infiniteCode has time to refactor? It is 8 lines of code that will cover the vast majority of jolokia configurations.

[sparrc]

sorry @ScubaDrew, but in my experience once it's in it will never go away, in addition to being harder to understand and harder to refactor in the future.

If someone else can take on properly refactoring and testing the code I'd appreciate it a lot. In the meantime I'm not going to merge in a hack.

[ScubaDrew]

@sparrc -- Understandable and thanks for the consideration. I have an enviornment running 1.0.0 which is suffering from this issue and thus is failing to collect many of the jolokia metrics.

[sparrc]

closing for lack of updates on requested changes, please feel free to resubmit a PR if you can address the requested changes